Socioeconomic and Environmental Dimensions of Agriculture, Livestock, and Fisheries: A Network Study on Carbon and Water Footprints in Global Food Trade

Abstract

Agriculture, livestock, and fisheries significantly impact socioeconomic, environmental, and health dimensions at global level, ensuring food supply for growing populations whilst promoting economic welfare through international trade, employment, and income. Considering that bilateral food exchanges between countries represent exchanges of natural resources involved in food production (i.e., food imports are equivalent to savings of natural resources), the purpose of the study is to investigate the evolution of carbon and water footprints corresponding to the global food trade networks between 1986 and 2020. The research aims to identify potential associations between carbon and water footprints embedded in food trade and countries’ economic welfare. Complex network analysis was used to map countries’ positions within annual food trade networks, and countries’ metrics within networks were used to identify connections between participation in global trade of carbon and water footprints and economic welfare. The findings of the study show an increase in carbon and water footprints linked to global food exchanges between countries during the period. Furthermore, a country’s centrality within the network was linked to economic welfare, showing that countries with higher imports of carbon and water through global food trade derive economic benefits from participating in global trade. Global efforts towards transformations of food systems should prioritize sustainable development standards to ensure continued access to healthy sustainable diets for populations worldwide.

1. Introduction

Agriculture, livestock, and fisheries are major sectors within food systems, encompassing complex interactions among numerous agents involved in food production, exchange, and consumption [1]. The identification of trends and determinants of food consumption patterns and food production capacity at national and international levels may support the establishment of positive dynamic interactions among agents within food systems, based on sustainable development standards [2], i.e., strategies towards controllable paths favoring environmental preservation to ensure human existence over time. However, individuals, organizations, and countries within food systems face multiple trade-offs linked to social, political, economic, and health dimensions, particularly concerning hunger, poverty, and unemployment [3,4,5,6], which represent interconnected issues permeating food production, trade, and consumption at local, regional, and global levels [7,8].

The concentration of food production in specific regions of the world imposes the intensification of global connectivity through bilateral trade to ensure adequate food supply for populations worldwide [9]. Simultaneously, food exchanges between countries within trade networks rely on extensive use of human, financial, and natural resources, exerting environmental pressures on national food systems. Thus, international trade may lead to social and economic pressures at the local level towards increased productivity, in addition to environmental pressures based on greenhouse gas emissions and depletion of natural resources that are transferred virtually within goods [10].

Mapping synergies and trade-offs within global food trade networks requires analyzing hidden costs linked to the production, trade, and consumption to address inequalities in the distribution of food, water, employment, and land [11]. Therefore, the purpose of the present study is to investigate the evolution of carbon and water footprints corresponding to global food trade networks between 1986 and 2020. The objective of the research is to identify potential effects on economic welfare arising from countries’ participation and position within global food trade networks.

The investigation addresses the following research question: Do higher imports of carbon and water footprints embedded in global food trade offer benefits in terms of higher economic growth due to potential savings of natural resources? Therefore, the present study contributes to the literature on sustainable development standards by providing evidence on the economic advantages linked to countries’ involvement in exchanges of carbon and water footprints transferred through global food trade networks.

2. Literature Review

Environmental pressures may be represented through diverse metrics referring to full life-cycle environmental responsibilities from production to the final consumption of products and services [12]. The complex interplay of international trade relationships and their impacts on domestic environmental degradation requires countries to acknowledge the effects originating throughout the entire supply chain to effectively address the sustainability issue, based on environmentally extended multiregional input–output analysis (EE-MRIO) [13]. However, current global supply chain databases may be inadequate to account for the specific environmental impacts of diverse food products and capture the physical basis of food systems [14]

The increasing displacement of environmental impacts from primary production through global trade has become a prominent issue in international policy debates [14]. Food exchanges through global trade represent displacements of environmental pressures from the country of final consumption to the country of production through land-use studies [12,15]. Initiatives towards investigating global displacement include the Global Footprint Network (GFN) and the Water Footprint Network (WFN), proposing the establishment of national-level footprint accounts and acknowledging that the “embodied impact” on the environment may represent an intangible measure that implies allocation of “responsibility” for tangible emissions to the products that are outputs of the production system [13]. Footprint indicators comprise useful metrics for building national and regional accounts on environmental impacts from human activities, representing the aggregated output corresponding to the combination of outputs from economic activities within their specific geographic boundaries [12,13,16].

Evidence shows that the dynamic nature of carbon emissions transfers between countries involved in global trade significantly influences countries’ carbon emission profiles. Furthermore, countries’ positions within trade networks substantially impact their direct carbon emissions; therefore, major importing countries should assume part of the responsibility for reducing environmental pressures of their trade partners. Whilst strategic relationships may be crucial in reducing domestic emissions, the practice of outsourcing emissions raises ethical questions considering the interconnectedness of global supply chains and the global nature of climate change impacts [17].

Food systems generate 20% to 30% of global carbon footprints [18,19], being particularly concentrated in the production of animal food products and rice [19]. Furthermore, greenhouse gas emissions from food production, trade, and consumption represent a higher burden in developing countries [20]. In contrast, food items and agricultural products represent approximately 43% and 35% of water footprints, respectively, although agricultural production is responsible for 69% of global water withdrawals [21]. However, some authors maintain that international trade may contribute to reducing environmental impacts through global value chains, ensuring the supply of energy and natural resources (including carbon and water footprints) from peripheral economies to high-income countries [22]. Furthermore, recent research emphasizes that food systems and climate changes present bidirectional causality through the occurrence of extreme weather events, indicating the possibility of tipping points in global food supply [19].

Yet, it is important to emphasize that the majority of studies exploring effects linked to global food networks focus on isolated analyses of health-related issues of food systems, i.e., food sovereignty and resilience to shocks [23,24,25,26,27,28], exposure and traceability of food poisoning outbreaks [29], or food security and nutrition [7,8,30,31,32,33]. Other studies investigate economic issues related to trade networks of certain agricultural commodities and agricultural products in general [34,35]. The literature on global food trade networks presents a gap in assessing multidimensional connections between socioeconomic and environmental dimensions involved in food production, exchange, and consumption [36,37]. Therefore, the present research proposes to fill the gap in the literature through the combination of network analyses and statistical modeling, contributing with evidence on the short- and long-term effects of carbon and water footprint exchanges on economic welfare.

3. Materials and Methods

3.1. Study Design

The study comprises an empirical investigation based on quantitative analysis of data referring to carbon and water footprints linked to bilateral food trade between pairs of reporting and partner countries from 1986 until 2020. Carbon and water footprints were estimated using publicly available data extracted from the Statistics Division of the Food and Agriculture Organization (FAOSTAT) and the World Integrated Trade Solution (WITS) platform from the World Bank. Furthermore, the study investigates associations among countries’ positions within carbon and water footprint networks, their globalization levels, sociodemographic and geographical characteristics, and economic welfare during the period.

The study’s hypothesis is that countries with higher imports of carbon and water footprints embedded in global food trade benefit from higher economic growth, potentially due to savings of natural resources. The study relies on a twofold strategy: network analyses of carbon and water footprints within food trade, and regression analyses of associations among countries’ characteristics and position within food trade networks (Figure 1). Therefore, the investigation was based on the development of two datasets:

• The first dataset, used for the network analysis, was based on bilateral trade data of carbon and water footprints embedded within food items between pairs of countries. The data reflects the import perspective (i.e., the reporter country is the importer, and the partner country is the exporter), organized by year.
• The second dataset, used for the regression analysis, included demographic, socioeconomic, political, geographical, and trade characteristics of countries, including globalization indices and countries’ positions within trade networks related to carbon and water footprint exchanges, organized by year.

3.2. Data Sources

Data downloaded from the following six sources were used to construct the two datasets for the study:

1.

Food and Agriculture Organization (FAOSTAT);

2.

World Bank:

• World Integrated Trade Solution (WITS);
• World Bank country and lending groups;
• World Development Indicators;

3.

Su-EATABLE LIFE (SEL) database;

4.

Konjunkturforschungsstelle Swiss Economic Institute (KOF);

5.

Economic Complexity Observatory;

6.

Atlas of Economic Complexity from the Growth Lab.

The first dataset, derived from FAOSTAT [38] and WITS [39], contained information on merchandise exchanges (food import flows) between pairs of countries from 1986 to 2020. Countries engaged in global food trade were categorized by income level according to the World Bank’s country and lending group classifications [40]. Subsequently, the food import flows were converted into aggregate carbon and water footprints using conversion factors per kilogram/liter of food commodity from the SEL database [41].

FAOSTAT trade data covers 440 agricultural and livestock products from approximately 270 countries, which are regularly compiled, standardized, and disseminated based on the Food and Agriculture Organization’s (FAO) standard International Merchandise Trade Statistics (IMTS) Methodology. The trade matrix data are organized in origin–destination matrices available for bulk download, facilitating analysis of trade links between countries from 1986 to 2020. Due to the lack of fishery-related trade data in FAOSTAT, data on 20 fisheries products from WITS (Supplementary Materials, Table S1) were incorporated into the first dataset for the corresponding period, following methods from previous studies [7,8,30]. WITS provides detailed data on various traded goods, including food, commodities, and industrial products, formatted in origin–destination matrices that are compatible with the FAOSTAT data structure. The World Bank’s income level classification was used to categorize countries into four levels: low income (L), lower-middle income (LM), upper-middle income (UM), and high income (H) [40].

The carbon (kg CO₂ eq) and water (liters H₂O) footprint data per kilogram (kg) or liter (L) of food commodity from the SEL database [41] is based on a compilation of footprint values from peer-reviewed and grey literature. Standardized methodologies were applied to ensure the consistency and accuracy of environmental resource data related to food production and distribution. The database includes carbon footprint values for 325 food items, and water footprint values for 321 food items, organized by aggregation level [41].

The second dataset contained countries’ network metrics related to global carbon and water footprint trade, derived from network analysis. Additional data on countries’ demographic, geographical, and socioeconomic characteristics were obtained from the World Bank’s World Development Indicators [42], the Konjunkturforschungsstelle Swiss Economic Institute’s KOF Globalisation Index (measuring social, economic, and political dimensions) [43], the Economic Complexity Observatory [44], and the Atlas of Economic Complexity from the Growth Lab [45].

The World Development Indicators platform provides 1509 series of demographic, social, economic, political, structural, health, and environmental characteristics for 266 countries from 1960 to 2024. The KOF Globalisation Index encompasses de facto and de jure data for three dimensions of globalization (social, economic, and political) for 215 countries and world regions between 1971 and 2022. De facto globalization indicators summarize metrics related to physical and digital flows of individuals, merchandise, and information to and from countries. De jure indicators relate to metrics on legal frameworks and institutional structures that support international relations, access to information, civil rights, and global trade. Additional data on countries’ economic complexity, reflecting knowledge intensity and diversity in productive capabilities, were obtained from the Observatory of Economic Complexity [44] and the Atlas of Economic Complexity [45], and included in the statistical analyses; however, the variable showed lack of significance in the regression models.

3.3. Variables

Bilateral food trade data from the FAOSTAT [38] and WITS [39] platforms (1986–2020) were converted to the net weight (in kilograms) of the edible portion of food items exchanged for human consumption using domestic supply-to-food ratios and technical conversion factors, following methods from previous studies [5,6]. The decision to focus on agricultural, livestock, and fisheries trade exclusively for human consumption reflects the substantial differences in technical standards based on the final destination of the food items. The calculation of carbon (kg CO₂ eq) and water (liters H₂O) footprints embedded in the edible portion of food items follows the general structure outlined in Equation (1):

$$F_{xijt}=E_{xijt}\times F_x$$

(1)

where F_xijt is the carbon or water footprint attributable to exchanges of food item x between country i and country j in year t, E_xijt is the net weight of edible portion of food item x exported from country i to country j in year t, and F_x is the carbon or water footprint per kilogram/liter of the food item x.

Footprints were calculated by multiplying the volume of food item exchanges by the corresponding carbon and water footprint coefficients in the SEL database [41], providing the carbon or water impact for each bilateral food flow. Footprints estimated for individual food items (1, …, n) were then summed to derive aggregate footprints exchanged between country i and country j in year t to represent the flow of carbon and water embedded in foods through global trade.

The selection of carbon and water coefficients from the SEL database prioritized the closest match between the detailed descriptions of food items in FAOSTAT/WITS and SEL. The coefficient for the specific item was adopted whenever possible (e.g., food items described as “frozen” were matched with coefficients for “frozen” items). Given the study’s focus on the local environmental effects from food production and distribution on national economic welfare, coefficients for “imported” food items were excluded from the carbon and water footprint estimations, primarily due to data scarcity and resulting issue of comparability with other coefficients [41].

The structure of the first dataset, derived from FAOSTAT, WITS, the World Bank, and the SEL database, includes the following variables:

• Country of origin (exporter);
• Income level of the country of origin (exporter);
• Country of destination (importer);
• Income level of the country of destination (importer);
• Year of trade transaction;
• Trade flows of carbon (kg CO₂ eq) or water (liters H₂O) footprints within foods.

Countries were categorized into five income levels (high income, upper-middle income, lower-middle income, low income, and undefined income) based on the World Bank’s country and lending group classifications, which corresponds to the annual gross national income (GNI) per capita, calculated using the Atlas method [40]. The first dataset was used to perform the network analyses (further details in Section 3.4).

Countries’ network metrics from the network analyses were included in the second dataset, which also contained demographic, socioeconomic, geographical, and globalization characteristics, organized by year (

Table 1. Characteristics of variables in the dataset for regression analysis.

Regression Analysis		N	μ	SD	Min	Max
Population ages 0–14 years old	(%)	7006	31.17	10.88	11.60	51.18
Population ages 15–64 years old	(%)	7006	61.64	7.22	46.10	86.08
Population ages ≥ 65 years old	(%)	7006	7.19	5.12	0.17	29.58
Population, male	(%)	7006	50.04	2.73	45.11	76.61
Population, female	(%)	7006	49.96	2.73	23.39	54.89
Urban population	(%)	7006	56.00	24.13	5.13	100.00
Labor force unemployment	(%)	3705	8.28	5.83	0.10	38.80
Life expectancy at birth	(years)	6691	68.29	9.71	14.10	85.39
Fertility rate	(per woman)	6695	3.24	1.72	0.79	8.79
GDP per capita, 2021 PPP	(ln)	5829	9.31	1.21	6.22	12.07
Food dependency ratio		4807	434.66	656.55	0.00	9637.45
Indegree centrality for CWF		6752	0.31	0.23	0.00	0.94
Outdegree centrality for CWF		6752	0.31	0.25	0.00	0.94
Betweenness centrality for CWF		6752	0.00	0.01	0.00	0.05
Economic globalization de facto		6400	54.64	17.49	6.75	98.85
Economic globalization de jure		6034	50.31	20.34	6.75	96.72
Social globalization de facto		6496	50.35	21.81	4.06	97.74
Social globalization de jure		6512	55.76	20.89	6.11	95.18
Political globalization de facto		6603	50.50	27.43	1.36	98.85
Political globalization de jure		6603	58.69	25.07	1.00	99.69
Region		7621	4.50	2.20	1.00	8.00
Year		7621	-	-	1986	2020

N = observations; μ = mean; SD = standard deviation; ln = natural logarithm; GDP = gross domestic product; PPP = purchase power parity; CWF = carbon and water footprints.

). The second dataset was used to estimate multiple linear regression models to investigate the short- and long-term effects of countries’ participation and position within global trade networks on their gross domestic product per capita (further details in Section 3.5).

3.4. Network Analyses

The network analyses established links to represent and measure properties of carbon and water footprints embedded in bilateral food trade connections between countries from 1986 to 2020. The links may be represented by an adjacency matrix (A) for global trade between countries i (1,…, p) and countries j (1,…, p) in period t with a main diagonal of 0 (i.e., there is absence of trade of the country with itself), as per Equation (2):

$$A_t=\left[\begin{array}{ccc}0& \dots & a_{1pt}\\ ⋮& \ddots & ⋮\\ a_{p1t}& \dots & 0\end{array}\right]$$

(2)

where a_ijt represents the aggregate flow of carbon or water footprints exchanged between country i and country j in year t, based on original FAO/WITS data (in kilograms/liters) converted to carbon and water footprints using SEL coefficients (in kg CO₂ eq or liters H₂O per kilogram/liter of food).

Therefore, annual networks representing local environmental impacts (carbon or water footprints) resulting from participation in global food trade were constructed for the period 1986–2020. The nodes of the networks represent countries, and the edges represent trade flows weighted by their respective carbon or water footprints.

Nodes were categorized by country income level, based on the World Bank classification [40]. The size of each node represents the degree of a country’s participation in the network, and the thickness of the edges represents the volume of carbon/water footprint flow between connected countries, enabling a synthetic graphical representation of data for the analysis of international trade flows between countries across diverse income classes. Data organization, processing, analysis, and network modeling were performed using Python 3 language, and networks graphs were generated using Gephi version 0.10.

Network metrics referring to average weighted degree, diameter, density, modularity, average clustering coefficient, and average path length were extracted from the network analysis to identify trends in the evolution of global food trade from 1986 to 2020 [46,47,48]. Network parameters were estimated by applying network analysis techniques to the bilateral food trade data (Equations (3)–(8)).

$${KW}_{nt}={\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\left[n_t\right]$}\right.}\times {\sum }_i{\sum }_j{w}_{ijt}$$

(3)

where KW_nt is the average weighted degree of the network n in period t, n_t represents the number of countries in the network in the year t, and w_ijt corresponds to the weighted connections between countries i and j in relation to other countries in period t.

$${DI}_{nt}=\underset{d}{\max }\left(d_{ijt}\right)$$

(4)

where DI_nt is the diameter of the network n in period t, and d_ijt refers to the shortest distance between any pair of countries i and j during period t. Therefore, the diameter of the network corresponds to the shortest distance between the two countries situated most far apart in the trade network in a certain year.

$${DE}_{nt}={\displaystyle \raisebox{1ex}{$m_t$}\!\left/ \!\raisebox{-1ex}{$\left[n_t\times \left(n_t-1\right)\right]$}\right.}$$

(5)

where DE_nt is the density of the network n in period t, m_t corresponds to the number of edges in the network in period t, and n_t was previously defined in Equation (3).

$$M_{nt}={\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\left(2m_t\right)$}\right.}\times \sum \left[a_{ijt}-\left(k_{it}\times k_{jt}\right)/\left(2m_t\right)\right]\times \delta \left(c_{it},c_{jt}\right)$$

(6)

where M_nt is the modularity score of the network n in period t, m_t and a_ijt were previously defined in Equations (2) and (5), k_it and k_jt refer to the degree of country i and country j (respectively), and δ(c_it, c_jt) represent the Kronecker delta function indicating whether country i and country j belong to the same community (1) or no (0) in a certain year.

$${CC}_{nt}=\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$n_t$}\right.}\right)\times \sum 2\times t_{it}/\left[k_{it}\times \left(k_{it}-1\right)\right]$$

(7)

where CC_nt is the average clustering coefficient of the network n in period t, n_t and k_it were previously defined in Equations (3) and (6), and t_it represent the number of triangles connected to country i in period t.

$$P_{nt}={\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\left[n_t\times \left(n_t-1\right)\right]$}\right.}\times \sum d_{ijt}$$

(8)

where P_nt is the average path length of the network n in period t, and n_t and d_ijt were previously defined in Equations (3) and (4).

3.5. Statistical Analyses

Countries’ network metrics extracted from the network analyses, organized by year, were incorporated into the second dataset to comprise the synthesis of countries’ positions in the global carbon and water footprint trade related to food exchanges. The second dataset comprises a panel of country-level data including variables on demographic, socioeconomic, geographical, and globalization characteristics.

Descriptive analysis (central tendency and variability) and multiple linear regression models were used to identify of potential effects of countries’ positions within the international carbon and water footprint trade network and their globalization levels on economic welfare, represented by the domestic gross product per capita in 2021 purchasing power parity (PPP).

Multiple linear regression models were estimated using a stepwise process for selection of independent variables potentially related to lagged dependent variables that correspond to economic welfare, considering zero, five, and eight years after capturing countries’ positions in network metrics and other country characteristics. The approach allows for the identification of short- and long-term effects of changes in country participation within the global trade network (Equation (9)).

$$\ln {GDP}_{it+z}={\beta }_0+{\beta }_1\times N_{it}+{\beta }_2\times G_{it}+{\beta }_3\times D_{it}+{\beta }_4\times S_{it}+{\beta }_5\times C_{it}+\epsilon$$

(9)

where ln GDP_it+z corresponds to the natural logarithm of the GDP per capita (in 2021 PPP) of country i in the period t + z (being z = 0, 5, or 8); N_it is the matrix of metrics of the country i within the global trade network of carbon and water footprints in period t; G_it is the matrix of globalization metrics in economic, social, and political dimensions of the country i in period t; D_it is the matrix of demographic characteristics of the country i in period t; S_it is the matrix of socioeconomic characteristics of the population of country i during period t; C_it is the matrix of control variables referring to region of country i, year t, and interaction between region and year; and ε = error term.

The choice of lag periods for the statistical analysis was based on previous studies indicating increased robustness of models using lagged dependent variables [49]. Specifically, interdisciplinary studies of sustainable environmental–economic growth and economic complexity have been based on lag lengths ranging from 0 to 8 years in their models [7,8,50,51]. Therefore, we tested lag lengths from 0 to 8 years for model selection, using the Schwarz information criterion.

Considering the lack of economic welfare data (the dependent variable in the regression models) before 1990, the statistical analyses are limited to information from 1990 onward. Post-estimation tests were performed to identify potential multicollinearity among variables included in the regression models, to avoid bias in the estimations. Variables exhibiting a high variance inflation factor (VIF) were excluded. The regression models also include control variables for region, year, and the interaction between region and year to account for potential trade policies and agreements established between countries from 1990 to 2020. The statistical analyses were conducted using software Stata version 17 with a significance level of p < 0.05.

4. Results

4.1. Network Analyses

The findings from the network analyses include data on 255 countries and territories (Supplementary Materials, Table S2) and cover carbon and water footprint exchanges related to 344 food items (Supplementary Materials, Table S3). The data indicate an increasing trend in carbon and water footprints flows associated with food trade over the study period. The density and average degree of the carbon and water footprints embedded within the global food trade network nearly tripled from 1986 to 2020, indicating increased connectivity between countries. The weighted degree for carbon and water footprints grew more than the density and average degree, indicating stronger connections between countries (

Table 2. Network metrics for carbon and water footprints linked to food trade according to year.

Metric	1986	1990	2000	2010	2020
Average degree	21.796	22.196	47.557	55.545	62.447
Weighted degree for carbon	0.253	0.228	0.558	0.773	1.036
Weighted degree for water	420.568	405.195	857.171	1123.743	1481.105
Network diameter	4.000	3.000	4.000	4.000	3.000
Density	0.086	0.087	0.187	0.219	0.246
Modularity	0.491	0.450	0.458	0.499	0.480
Average clustering coefficient	0.477	0.548	0.579	0.591	0.620
Average path length	1.714	1.645	1.695	1.691	1.636

).

Yet, the average clustering coefficient showed only a minor increase over the period, suggesting that the density of links within countries’ neighborhoods exhibited moderate strength in forming groups. Additionally, the slight decrease in modularity and average path length combined with the stability in network diameter suggests a low latency in establishing connections between countries, i.e., countries within the network maintained a relatively strong community structure throughout the period of analysis.

Graphs for 1986, 2000, and 2020 illustrate networks representing carbon and water footprint exchanges in global food trade, showing the increasing growing trends in environmental footprints associated with bilateral food trade. Additionally, the graphs highlight the dominant role of high-income countries, representing of proportional sizes based on their degrees within the global networks (Figure 2).

The evolution of carbon and water footprint exchanges in global food networks shows an increasing prominence of France and the Netherlands during the period of analysis (Figure 2). Figure 3 generalizes the data from Figure 2, depicting network graphs that aggregate carbon and water footprint exchanges based on countries’ income levels. General trends reveal a consistent maintenance of trade patterns from 1986 to 2020. The highest proportion of exchanges occurred within groups of high-income countries with a decreasing proportion of connections among upper-middle-, lower-middle- and low-income countries. Furthermore, exchanges to and from countries across diverse income levels exhibited similar patterns, with major connections to high-income countries, followed by upper-middle-income countries. Conversely, lower-middle and low-income countries showed minimal participation in trade networks (Figure 3).

Analysis of carbon and water footprint exchanges over the study period, categorized by country groups, reveals slight reductions in the participation of upper-middle- and lower-middle-income countries in relation to high-income countries (Figure 3). The participation of high-income countries in global exchanges of carbon and water footprints linked to food trade decreased from 1986 to 2020, although their share in global markets still represents a significant portion of the overall footprint volume. Upper-middle-income countries experienced a substantial increase in participation over the period; whereas lower-middle-income countries showed only minor increases in carbon footprint imports, and water footprint exports and imports. In contrast, low-income countries and countries with undefined income level presented a considerable decrease in participation (

Table 3. Participation in carbon and water footprints linked to food trade according to income level and year.

Proportion of Trade Volume (%)	1986	1990	2000	2010	2020
Carbon footprint exports
High income	68.87	67.93	66.67	63.87	59.26
Upper-middle income	9.39	8.89	15.28	26.16	28.07
Lower-middle income	12.99	17.43	10.12	9.09	12.05
Low income	5.55	5.64	7.91	0.84	0.57
Undefined income level	3.20	0.11	0.02	0.03	0.06
Carbon footprint imports
High income	65.62	68.15	63.41	63.71	56.74
Upper-middle income	11.25	13.66	13.95	25.61	29.30
Lower-middle income	9.57	9.11	17.48	9.04	11.84
Low income	3.25	9.08	5.16	1.62	1.92
Undefined income level	10.31	0.00	0.00	0.02	0.20
Water footprint exports
High income	56.42	56.29	56.94	54.99	50.75
Upper-middle income	11.60	10.03	16.67	29.33	31.00
Lower-middle income	19.05	23.16	12.50	13.60	16.68
Low income	11.09	10.47	13.88	2.05	1.53
Undefined income level	1.84	0.06	0.02	0.03	0.04
Water footprint imports
High income	62.49	66.83	61.75	59.68	51.33
Upper-middle income	10.81	11.24	13.73	28.76	32.37
Lower-middle income	10.15	9.84	18.03	9.65	13.91
Low income	4.11	12.09	6.49	1.89	2.20
Undefined income level	12.44	0.00	0.00	0.02	0.20

).

4.2. Statistical Analyses

Demographic trends of countries indicate an ongoing demographic transition, characterized by population aging and increasing life expectancy at birth. Additionally, there was an increase in the proportion of the global urban population, and a rise in the food dependency ratio, indicating a growing reliance on food imports throughout the period.

The indegree and outdegree centrality of countries involved in carbon and water footprint exchanges within global food trade networks (representing connections related to imports and exports, respectively), along with major part of globalization indices, show increase from 1990 to 2020, whereas betweenness centrality maintained stable trends (

Table 4. Demographic, socioeconomic, globalization, and trade characteristics of countries.

Characteristics		1990		2000		2010		2020
		μ	SD	μ	SD	μ	SD	μ	SD
Population ages 0–14 years old	(%)	35.857	10.106	32.312	10.342	28.547	10.718	26.918	10.476
Population ages 15–64 years old	(%)	58.253	6.892	60.986	6.814	63.813	7.211	63.767	6.264
Population ages ≥ 65 years old	(%)	5.890	4.128	6.702	4.588	7.640	5.302	9.315	6.469
Population, male	(%)	50.191	2.433	49.958	2.496	50.032	3.049	50.044	2.962
Population, female	(%)	49.809	2.433	50.042	2.496	49.968	3.049	49.956	2.962
Urban population	(%)	51.125	25.420	55.055	23.816	58.419	23.509	61.057	23.172
Labor force unemployment	(%)	7.611	5.214	8.750	6.071	8.476	6.159	7.807	5.136
Life expectancy at birth	(yrs)	64.405	10.187	67.116	9.813	70.452	8.824	72.179	7.522
Fertility rate	(pw)	4.148	1.901	3.222	1.722	2.902	1.511	2.559	1.291
GDP per capita, 2021 PPP	(ln)	8.999	1.214	9.161	1.224	9.446	1.189	9.537	1.149
Food dependency ratio		300.179	441.760	399.237	601.989	430.212	669.730	439.880	602.435
Indegree centrality for CWF		0.185	0.128	0.302	0.215	0.348	0.239	0.413	0.271
Outdegree centrality for CWF		0.185	0.246	0.302	0.235	0.348	0.238	0.413	0.235
Betweenness centrality for CWF		0.002	0.004	0.003	0.006	0.003	0.005	0.003	0.005
Economic globalization de facto		47.330	17.707	54.858	16.181	58.397	16.696	58.196	17.267
Economic globalization de jure		39.810	18.239	51.615	21.989	55.292	18.723	56.053	19.200
Social globalization de facto		40.019	20.774	45.285	20.964	58.123	19.808	60.246	19.045
Social globalization de jure		42.901	20.045	52.568	20.117	63.845	17.878	67.270	16.675
Political globalization de facto		46.737	25.973	48.391	28.169	54.005	27.306	56.551	26.519
Political globalization de jure		39.855	19.196	57.937	23.320	66.931	23.191	70.387	22.552

μ = mean; SD = standard deviation; yrs = years; pw = per woman; ln = natural logarithm; GDP = gross domestic product; PPP = purchase power parity; CWF = carbon and water footprints.

).

The coefficients of the regression models suggest that the indegree centrality and betweenness centrality for countries participating in carbon and water footprint exchanges within global food trade networks are positively associated with GDP per capita, i.e., higher imports of carbon and water footprints embedded in food and higher connectedness within trade networks, are linked to higher income per capita. Furthermore, the associations identified for countries’ network metrics in carbon and water footprint exchanges remained consistent over time, presenting higher coefficients compared to other variables in the models (

Table 5. Coefficients of regression models for economic growth in relation to exchanges of carbon and water footprints linked to food trade.

GDP Per Capita, 2021 PPP (ln) §	t0			t5			t8
	β	SE	Sig.	β	SE	Sig.	β	SE	Sig.
Population ages 15–64 years old	0.0283	0.0017	***	0.0230	0.0026	***	0.0228	0.0029	***
Population, male	0.0308	0.0029	***	0.0374	0.0046	***	0.0416	0.0051	***
Urban population	0.0089	0.0004	***	0.0093	0.0007	***	0.0096	0.0008	***
Labor force unemployment	−0.0072	0.0011	***	−0.0067	0.0017	***	−0.0054	0.0019	**
Life expectancy at birth	0.0111	0.0022	***	0.0131	0.0034	***	0.0166	0.0037	***
Food dependency ratio	−0.0002	0.0000	***	−0.0002	0.0000	***	−0.0002	0.0000	***
Indegree centrality for CWF	0.1222	0.0565	*	0.4003	0.0875	***	0.2160	0.0939	*
Outdegree centrality for CWF	−0.0726	0.0661		−0.0033	0.1046		−0.0004	0.1126
Betweenness centrality for CWF	11.0370	1.8400	***	7.9669	2.7890	**	10.9553	2.9934	***
Economic globalization de facto	−0.0024	0.0006	***	−0.0014	0.0009		−0.0002	0.0010
Economic globalization de jure	0.0007	0.0006		0.0002	0.0009		−0.0013	0.0010
Social globalization de facto	0.0218	0.0009	***	0.0183	0.0014	***	0.0192	0.0015	***
Social globalization de jure	0.0134	0.0010	***	0.0150	0.0015	***	0.0123	0.0016	***
Political globalization de facto	−0.0009	0.0006		−0.0007	0.0010		−0.0012	0.0011
Political globalization de jure	0.0009	0.0008		−0.0026	0.0013	*	−0.0034	0.0014	*

β = coefficient; SE = standard error; GDP = gross domestic product; PPP = purchase power parity; ln = natural logarithm; CWF = carbon and water footprints. ^§ Models include control variables for region, year, and crossed effects between region and year. * p < 0.05; ** p < 0.01; *** p < 0.001.

).

5. Discussion

The study’s findings indicate that the evolution of global food trade networks was characterized by an intensification of bilateral exchanges between countries over the period of 35 years from 1986 to 2020, including substantial rise in carbon and water footprints embedded in food flows (Table 2). Additionally, the results supported the research hypothesis, showing that countries with higher imports of carbon and water footprints embedded in global food trade experience higher economic growth (Table 5), potentially due to the savings of natural resources. Furthermore, the network analyses show moderate strength in the formation of groups within the network (Table 2), which contrasts with evidence from a previous study using multilevel networks to identify communities of countries within global food networks [26].

Torreggiani et al. [26] showed densely connected groups within layers of specific food products; however, their network analysis focused on 178 countries trading 16 agricultural commodities representing staple foods with major contribution to calorie exchange at the global level between 1992 and 2011. In addition to differences in the scope of investigation (food items, countries, and period of analysis), their results incorporated crops primarily intended for animal consumption (e.g., soybeans, maize, sorghum, and barley) due to the lack of consideration for domestic supply-to-food ratios. Similarly, the use of multilevel networks may have captured differences in the structure of demand between food and feed, highlighting similarities related to geographical proximity and cooperation agreements among countries within regions, rather than economic similarities. Nevertheless, the authors identified low variability in certain features of food trade networks [26], consistent with our network analyses, which indicated low variation in modularity, average path length, and network diameter.

Additionally, our regression models indicated that a country’s strategic positioning within the global trade network (betweenness centrality), particularly associated with higher imports of carbon and water footprints related to food exchanges, was linked to greater economic welfare (Table 5), although lacking significance in the connections with countries’ economic complexity. The relationship between economic complexity, economic growth, and the environmental sustainability of countries garnered significant attention in the literature over the past decade [52,53,54,55,56]. Recent studies suggest that greater economic complexity is associated with lower greenhouse gas emissions per capita, particularly due to the adoption of innovative technologies and investments in the production of high-value-added goods, which may contribute to higher efficiency in production processes and a reduction in environmental impacts [55].

Other studies emphasize that greater economic complexity and growth may lead to improved environmental outcomes in certain situations, particularly through the adoption of advanced technologies supported by robust institutions [54,57,58,59,60]. In addition, certain studies suggest that economic complexity and growth may mitigate environmental degradation, especially after reaching certain thresholds [52,54]. However, many authors contend that economic complexity is positively correlated with the ecological footprints of production, i.e., whilst complexity convey technological advances, it also increases the environmental burden of production activities [53,56].

The evidence from our study suggests alternative approaches to the issue. The use of resources for producing high-value-added goods may be accomplished at the expense of production in agriculture, livestock, and fisheries, which may be sourced through global food trade at lower prices due to their nature of low-value-added commodities. Furthermore, food imports may represent the outsourcing of carbon and water footprints to other countries, supporting the hypothesis from previous studies that global trade creates intricate networks of ecological exchanges, allowing high-income/developed countries to shift the environmental burden of their consumption to low-income/developing nations, thereby increasing health and environmental risks in the latter [61,62].

The unequal exchange of environmental footprints is particularly evident in trade exchanges between low- and middle-income countries compared to high-income nations [63]. The food exchange patterns identified in our study reinforce findings from prior research, demonstrating negligible participation by low-income countries throughout the period (Figure 3 and Table 3). Furthermore, a significant portion of trade registered for low-income countries consists of food aid [64]. Moran et al. [61] showed that high- and middle-income countries tend to trade ecological footprints primarily among themselves with significant interregional flows, particularly from Latin America to North America and from North America to Asia–Pacific.

Marshall et al. [65] address the complexity of global food systems by proposing four key indicators related to supply chains, food environments, and patterns of agricultural consumption and production to categorize countries. Their findings indicate that industrialized food systems ensure higher agricultural productivity and food security for their populations, although they face significant environmental challenges; whilst traditional food systems tend to exhibit higher levels of undernutrition. Other authors highlight that high-income countries typically experience ecological impacts related to urbanization and consumption, whereas low- and middle-income countries bear the burden of production-related impacts through investments in exports [58,66].

Our results indicate a significant rise in the density of carbon and water footprints exchanged through global food networks (Table 2), highlighting the growing food demand from populations worldwide. The consumption patterns and production activities of countries may influence global productivity with direct implications for sustainability and resource management [61]. Willett et al. [67] emphasize the need for major transformations within global food systems, given their significant role in generating environmental footprints. Thus, our analysis of ecological footprints in international trade provides key evidence for identifying the global distribution of environmental impacts.

Policy recommendations related to food production and consumption patterns must be highlighted within the context of the present study. Transforming food systems towards environmentally sustainable production requires investments in human capital, and renewable energy to support the reduction of ecological footprints [52,54,58]. Janssen et al. [68] show that global food trade represents an important adaptive tool for minimizing the risk of hunger due to climate change. Furthermore, Geyik et al. [69] emphasize that sustainable food systems should address malnutrition and climate change in decisions about food production, focusing primarily on plant-based diets. Several authors indicate that targeting improvements in agricultural and livestock productivity [69], along with reductions in food losses and food waste [69,70,71], would allow for fulfilling the nutritional needs of the global population with only minor changes in food consumption patterns and minimizing environmental impacts.

Recent evidence underscores the substantial role of dietary choices in shaping environmental footprints, particularly regarding greenhouse gas emissions (GHGE) and water scarcity [72,73,74]. Yet, improvements in one sustainability indicator (e.g., carbon footprint) may inadvertently worsen another indicator (e.g., water footprint), highlighting the need for integrated sustainability assessments to navigate the complexity of public policy decision-making processes. While healthier diets generally exhibit lower GHGE, they may paradoxically increase water scarcity footprints due to the substitution of animal-source foods with plant-based foods, which require greater water usage [74]. Therefore, national nutrition data systems that integrate food composition databases with environmental and economic indicators are essential for identifying environmental and economic trade-offs associated with dietary patterns [75].

Additionally, social and health dimensions, particularly long-term patterns of human behavior and connections between diet and chronic diseases, are critical elements in policies promoting sustainable, healthier diets [72]. Although diets with a significantly lower carbon footprint may be affordable to a population due to increased competition from recent globalization trends [76], their acceptability may comprise a barrier to extensive adoption, underscoring the complex interaction between consumer preferences, dietary habits, and sustainability [73]. Furthermore, various aspects of international relations established between countries involved in global food trade may have conflicting effects on economic, environmental, and health dimensions at national and international levels [7,8,77,78].

The study presents certain limitations. Firstly, the data used to estimate carbon and water footprints within global food trade networks rely on information publicly available from multiple platforms of international organizations and research groups. While efforts have been made toward assembling, standardizing, and validating the data, we acknowledge potential issues stemming from missing data (i.e., a lack of complete information on trade flows, the absence of identifiable food trade partners, or low specificity in food item descriptions). Part of the limitations were addressed through methods applied for verification of the symmetry in food trade between import and export records [7,8], and standardization of procedures for matching and application of environmental footprints to food exchanges recorded from 1986 to 2020.

Secondly, the study specifically focuses on foods intended for human consumption, based on the use of domestic supply-to-food ratios calculated using food balance sheets from FAOSTAT. Yet, the ultimate domestic destination of imported foods may differ from that of domestic production. However, considering the lack of reliable information on the proportion of imported foods by destination type within countries, it is reasonable to assume a similarity between the actual destination of imported food flows and ratios estimated using comprehensive data from the food balance sheets.

Thirdly, the footprint indicators used in the study represent conceptual constructs within evolving global food systems. Our analyses focused on environmental footprint flows associated with food trade between countries, encompassing food supply measures. However, countries’ relative positions, referring to net imports or exports of carbon and water footprints, and expressed through food demand patterns, involve decisions detached from the corresponding embodied impacts on local production systems, and subjective values related to food production and consumption strategies [13].

Therefore, the study lacks identification of the economic, cultural, and social drivers of food demand, or inequities in distribution of environmental footprints within countries. Beyond the immediate environmental impacts of production and consumption, future research should examine the potential determinants of food demand and its unequal distribution among individuals in diverse populations. Nonetheless, the use of country-level data in the investigation provides a significant contribution to the evidence on the production and distribution of environmental footprints related to global food trade.

Fourth, the regression models in our statistical analyses employed ecological approaches to identify associations between metrics of environmental footprints embedded in food trade networks and economic welfare, i.e., findings should be interpreted cautiously due to potential confounding effects. Although our models include key determinants of GDP per capita based on open-source data from reputable institutions, limitations in the time series data prevented the inclusion of potential confounding variables at the national level (e.g., educational attainment, and infrastructure indicators). Nevertheless, the use of robust empirical strategies to control for other potential confounders and test for possible sources of bias in the estimates likely reduced effects attributable to unobservable variables.

Finally, it is important to note that information on economic welfare was only available from 1990 onward; therefore, regression models using lagged GDP per capita for 0, 5, and 8 years included data corresponding to panels of 30, 25, and 22 years, respectively. However, the findings of regression models showed short- and medium-term positive associations between indegree and betweenness metrics related to environmental footprints within the food trade network and economic welfare, pointing to the consistency of the coefficients estimated in the statistical analyses.

The study contributes to the existing literature by emphasizing the connections between environmental footprints embedded in global food trade networks and the economic welfare of countries involved in food exchanges. Our findings may contribute to identify priorities for transforming food systems towards greater sustainability at the global level. Furthermore, evidence from the study may comprise important inputs to inform policy strategies aimed at promoting economic welfare at the national level, particularly considering findings from previous investigations on nutrition and health outcomes linked to global food trade networks [7,8].

The findings from the model proposed in the study may be used to support advances in standardization practices adopted within food production and trade, thereby enabling further reductions in carbon and water footprints throughout national food systems. Furthermore, long-term planning strategies focused on economic growth should prioritize mapping activities that may contribute to improving standards for optimizing food production, e.g., innovative technologies for food production like precision agriculture and livestock farming, vertical farming, and others [78,79,80,81]. Considering the significant effects of intermediation (betweenness centrality) in carbon and water footprint exchanges on economic growth, countries assuming leadership in innovation related to food production and trade policies within their regions may experience additional economic benefits.

Furthermore, changes in consumer preferences play an important role in transforming food systems, particularly in the context of the post-COVID-19 pandemic. Recent trends towards healthier food consumption patterns like local production, organic foods, and plant-based diets may present challenges to existing standards in food production [82,83]. The major part of consumers primarily focuses on price, taste, and perceived health benefits of foods [84], often associating organic foods with superior health properties [82]. While plant-based foods may contribute to greater food system sustainability by reducing greenhouse gas emissions, land use, and biodiversity losses, they also lack nutritional diversity and may generate higher water and energy consumption depending on the specific crops selected [85,86,87].

Whilst recent trends in food production and consumption show substantial changes, it is important to acknowledge that innovations have primarily been adopted in high-income countries, and may yield important economics–health–sustainability trade-offs (i.e., certain technologies may reduce carbon footprints but increase water footprints, while some innovations may increase food sovereignty but decrease trade at increasing costs and unemployment, etc.). Therefore, future research should focus on the mechanisms underlying the dynamics of decision-making processes related to food production, in addition to the factors that influence food consumption patterns and distributional issues at the national and local levels, which impact environmental footprints connected to global food trade networks. Although publicly available datasets present limitations regarding elements linked to the dynamics of individual food production and consumption choices, agent-based models and system dynamics modeling offer alternative tools for exploring features that capture the complexity of food systems [88].

6. Conclusions

The study’s findings contribute to the body of evidence regarding the economic advantages associated with countries’ involvement in exchanges of carbon and water footprints embedded in global food trade networks. Trends in the evolution of carbon and water footprints embedded in global food trade networks demonstrate an increase in environmental impacts from 1986 to 2020, reflecting the changing food requirements of a growing population. Yet, given the links established between economic welfare and networks position (referring to imports and intermediation of environmental footprint exchanges), national governments should carefully consider the potential consequences of policy strategies related to food production, trade, and consumption to identify best choices that balance environmental, health, and economic priorities. Furthermore, global efforts to transform food systems should prioritize sustainable development standards to ensure continuing access to healthy sustainable diets for populations worldwide. Further developments in the model proposed in the research will focus on the dynamics of food systems and potential sources of instability (tipping points) that may lead to either food insecurity and famine, or improvements in standards of food production and trade at national and global level, contributing to the sustainable transformation of food systems.